Controlling for baseline telomere length biases estimates of the rate of telomere attrition

Abstract

Longitudinal studies have sought to establish whether environmental exposures such as smoking accelerate the attrition of individuals' telomeres over time. These studies typically control for baseline telomere length (TL) by including it as a covariate in statistical models. However, baseline TL also differs between smokers and non-smokers, and telomere attrition is spuriously linked to baseline TL via measurement error and regression to the mean. Using simulated datasets, we show that controlling for baseline TL overestimates the true effect of smoking on telomere attrition. This bias increases with increasing telomere measurement error and increasing difference in baseline TL between smokers and non-smokers. Using a meta-analysis of longitudinal datasets, we show that as predicted, the estimated difference in telomere attrition between smokers and non-smokers is greater when statistical models control for baseline TL than when they do not, and the size of the discrepancy is positively correlated with measurement error. The bias we describe is not specific to smoking and also applies to other exposures. We conclude that to avoid invalid inference, models of telomere attrition should not control for baseline TL by including it as a covariate. Many claims of accelerated telomere attrition in individuals exposed to adversity need to be re-assessed.

Consequences of measurement error in qPCR telomere data: A simulation study

“…Those studies that have a high correlation between time 1 and time 2 TS ratio (which were non-qPCR studies) show a negligible dependency of TS ratio change on time 1 TS ratio, whereas those with a low correlation between time 1 and time 2 show a strong negative dependency. The empirical datasets do not align exactly with the simulation predictions, but some imprecision in the estimation of correlations should be expected since the empirical datasets have modest sample sizes (47–539, see [27]). Thus, one interpretation of these data is that some of the qPCR studies feature a high degree of measurement error, and show the predicted combination of low time 1-time 2 correlation and apparent negative dependency of the rate of change on the time 1 telomere length.…”

“…It is not a consequence of the TS ratio in particular, but of any set of repeated measurements where there is measurement error. An implication is that because at least part of the apparent association of initial telomere length and subsequent change is spuriously created by measurement error, controlling for initial telomere length in regression models in which the outcome variable is telomere length change is often invalid and biases inferences [27]. Specifically, it biases the estimate of the effect on telomere length change of any predictor variable that is associated with telomere length at baseline.…”

“…We then compared the patterns in the repeated samples simulations to the variation seen across longitudinal studies of human telomere dynamics (dataset 2). Dataset 2 comes from a collation of seven human cohort studies in which telomere length had been measured twice in the same individuals, an average of 8.5 years apart (range 6.0 to 9.5 years; see [27] for full details). Five studies used qPCR, and two measured terminal restriction fragment using Southern blot.…”

“…Thus, the extent of measurement error is likely to have varied. Of the various data collated in [27], we extracted the correlation coefficient between the time 1 and time 2 telomere length measurement, and the correlation coefficient between the time 1 measurement and the change in telomere length between time 1 and time 2, for comparison against simulated datasets with varying levels of measurement error.…”

Consequences of measurement error in qPCR telomere data: A simulation study

“…The second dataset (dataset 2) comes from a recent paper on longitudinal studies of human telomeres [27]. Bateson, Eisenberg and Nettle collated data from seven human cohort studies in which telomere length had been measured twice in the same individuals, an average of 8.5 years apart (range 6.0 to 9.5 years; see [27] for full details). Five studies used qPCR, and two measured terminal restriction fragment using Southern blot.…”